Every thing linked to the generative AI world strikes quick, together with AI brokers. Take, for instance, the Mannequin Context Protocol (MCP) that brokers use to entry information and instruments. The adoption fee of MCP is off the charts.
However can the brand new Agentic AI Basis (AAIF) transfer shortly sufficient to standardize your entire agentic AI ecosystem, beginning with MCP?
The thrill on the current MCP Dev Summit exhibits they’re off to a quick begin, however important challenges stay for enterprise adoption with safety as the highest concern.
The AAIF’s mission is to outline the open requirements for the agentic AI house “that the entire world will have the ability to undertake,” AAIF Government Director Mazen Gilbert instructed SD Occasions. “While you transfer from experimentation to manufacturing, and you’ve got a thousand individuals doing their very own factor and exhibiting you some cool demos and transferring to manufacturing at scale, you do want an open normal. Nobody firm can outline it, and nobody firm can personal it,” Gilbert added.
However can the AAIF transfer quick sufficient to stop trade fragmentation? Many organizations and software program corporations are already resolving key points with MCP and the broader agentic ecosystem exterior of the AAIF.
Quick Adoption and Quick Development
AAIF has grown to 170 members because it was launched in December, and attracted 1,100 attendees to its current MCP Dev Summit in New York. “Ninety-seven new members joined in February alone,” stated David Nalley of AWS, AAIF Governing Board Chair in his keynote. “AAIF already has seven working teams with greater than 5 hundred members,” he added. “The Id and Belief Working Group has 150 members from 66 organizations.”
At a media roundtable with main MCP maintainers from Anthropic, AWS, Microsoft, and OpenAI, panelists stated that transferring the specification to AAIF hasn’t modified a lot about their work course of and that it’s useful to have the broader perspective the AAIF members carry.
However MCP by itself represents solely part of the agentic AI panorama.
“MCP is the seed,” stated MCP maintainer Nick Cooper, Technical Employees, OpenAI on the media roundtable. “The inspiration has a broad mandate past simply MCP.”
And even throughout the scope of the MCP specification itself, organizations are expressing key necessities for enterprise adoption, reminiscent of reliability, scalability, governance, and enhanced safety. The maintainers stated they’re nicely conscious of enterprise necessities, and AAIF is requesting proposals to sponsor working teams that deal with them. For instance, enhanced safety enchancment is a key requirement for enterprise adoption, stated MCP maintainer David Soria Parra, Technical Employees, Anthropic, and co-creator of MCP. “We’re blissful to have AAIF carry the trade collectively and speak about the correct options” throughout the agentic ecosystem.
Key Messages from Keynotes
In his keynote, Linux Basis CEO Jim Zemlin talked in regards to the trade’s speedy embrace of AAIF and introduced a world occasion collection for the AAIF, in addition to expanded AgentCon + MCP Con occasions in Europe and North America in 2026. (The AAIF is a sub-foundation of the Linux Basis.) Zemlin additionally launched Angie Jones as the brand new Developer Expertise lead for AAIF
Nalley in his keynote additionally offered the just lately authorized AAIF venture life cycle (for standardization and introduced that AAIF is soliciting new venture proposals.
In Parra’s keynote he stated they’re monitoring greater than 110M SDK downloads per 30 days for MCP, and famous that it took React round three years to attain numbers much like what MCP has achieved in 16 months. He stated he’s seeing large adoption inside enterprise firewalls.
The MCP specification roadmap contains the next objects, Parra stated:
- Stateless transport — changing stateful server-sent occasions (SSE) to facilitate hyperscaler-friendly deployments
- Process primitive — to assist asynchronous and long-running agent operations
- Enterprise authentication — cross-org token trade for seamless single signal on (SSO)
- Triggers — server-initiated interactions (for instance, “I’ve new information, come have interaction”)
- Expertise — bundling domain-specific data with MCP servers
- SDK v2 for Python and TypeScript — that includes ergonomic cleanup and migration instruments
Parra additionally stated that the MCP working group is addressing the next agentic ecosystem challenges:
- Context bloat — utilizing progressive discovery and power seek for deferred loading
- Composability — structured outputs that permit fashions chain software calls with out round-tripping to inference
